Anda di halaman 1dari 14

Soils and Foundations 2014;54(6):11451158

HOSTED BY The Japanese Geotechnical Society

Soils and Foundations
journal homepage:

Pullout resistance of granular anchors in clay for undrained condition$

B.C. OKellya,n, R.B.J. Brinkgreveb,c, V. Sivakumard
Department of Civil, Structural and Environmental Engineering, Trinity College Dublin, Dublin, Ireland
Geo-Engineering Section, Delft University of Technology, CN Delft, The Netherlands
Research and Projects, Plaxis bv, Computerlaan 14, 2628 XK Delft, The Netherlands
Geotechnical Engineering SPACE, Queens University Belfast, Belfast, UK
Corresponding author. Tel.: 353 1896 2387; fax: 353 1677 3072.
Received 21 January 2013; received in revised form 2 August 2014; accepted 7 September 2014
Available online 13 December 2014


Granular anchors (GAs) can resist pullout/uplift forces, compression forces and also provide ground improvement. Under pullout loading,
a centrally located tendon transmits the applied surface load to the base of the granular column via a base plate attachment, which compresses the
column causing signicant dilation of the granular material to occur, thereby forming the anchor. This paper describes a program of eld testing
and numerical modelling of the pullout resistance of GA installations in overconsolidated clay for the undrained (short term) condition. Pertinent
modes of failure are identied for different column length to diameter (L=D) ratios. The applied pullout load is resisted in shaft capacity for short
GAs or in end-bulging of the granular column for long GAs. In other words, the failure mode is dependent on the column L=D ratio. A novel
modication in which the conventional at base-plate is replaced by a suction cup was shown to signicantly improve the undrained ultimate
pullout capacity of short GAs.
& 2014 The Japanese Geotechnical Society. Production and hosting by Elsevier B.V. All rights reserved.

Keywords: Bulging capacity; Failure; Granular anchor; Uplift; Ultimate capacity

1. Introduction expansive clays (Phanikumar et al., 2004, 2008; Sharma et al.,

2004; Srirama Rao et al., 2007). Another recent development
Granular anchors are a relatively new and promising is the jet mixing anchor pile, a supporting technology
foundation solution, particularly suited for lightly loaded particularly suited for foundation pit engineering in soft clay.
structures. In addition to the improvement provided to the The ultimate capacity and loaddeformation relationship of
surrounding ground, granular anchors can resist both pullout/ such piles have been investigated by Xu et al. (2014) using
uplift forces and compression forces. Hence they have been uplift eld tests and numerical analyses.
adopted, for instance, to prevent foundation uplift caused by The focus of the present study is to investigate the ultimate
ooding (Liu et al., 2006) or to resist foundation heave in capacity and loaddeformation relationship of granular anchor
(GA) foundations under uplift loading. The GA consists of
three main components (Fig. 1): a horizontal base plate, a
E-mail addresses: (B.C. OKelly), (R.B.J. Brinkgreve),
central vertical tendon (metallic rod or stretched cable) and (V. Sivakumar). densied granular material introduced into the borehole to
Peer review under responsibility of The Japanese Geotechnical Society. form a granular column. Under an applied uplift force (P), the
0038-0806/& 2014 The Japanese Geotechnical Society. Production and hosting by Elsevier B.V. All rights reserved.
1146 B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158

and resisted by the granular column. The dilatency of the

P granular material is a signicant factor controlling the GAs
pullout capacity. Recent experimental studies by OKelly et al.
(2013) and Sivakumar et al. (2013), among others, indicate
that the applied pullout load is resisted in shaft capacity for
short GAs or in localized bulging near the column base for
In-situ soil
long GAs. In other words, the failure mode depends on the
column length to diameter (L=D) ratio.
The motivations for the experimental and numerical studies
Tendon presented in this paper were to: (a) investigate the operation of
GAs, particularly the development of the pullout loaddis-
placement response for the undrained (short term) condition;
Granular (b) conrm the postulated modes of failure in shaft capacity or
column in end bulging and their dependence on the column L=D ratio
and ground conditions/properties; (c) develop appropriate
methods of analyses for the determination of the ultimate
Baseplate pullout capacity. The research programme involved perform-
ing 8 instrumented GA eld tests which were subsequently
modeled using nite element software. A novel modication of
the GA arrangement to improve its undrained ultimate pullout
Fig. 1. Schematic of granular anchor.
capacity was also modeled numerically.

tendon transmits the load to the column base via the base plate 2. Experimental programme
attachment. The resulting upward pressure over the column base
compresses the laterally conned granular column against the 2.1. Ground conditions
sidewall of the soil bore, thereby mobilizing an anchor resistance.
Unlike a conventional concrete anchor cast in-situ, pullout Full-scale eld trials were performed on 8 GAs installed in
loading can be applied to the GA immediately after its the upper Brown Dublin Boulder Clay (BrDBC) layer of the
installation. Signicant yielding occurs under pullout loading. Dublin Boulder Clay (DBC) deposit; an intact lodgement till.
For short GAs, this is also accompanied by signicant ground This is the primary supercial deposit within the greater
heave. In contrast, conventional concrete anchors generally fail Dublin region, Ireland. The DBC deposit is heavily over-
by sudden pullout on mobilizing the full shaft capacity, consolidated (it was deposited under ice sheets more than 1 km
assuming the anchor itself remains structurally sound. The in thickness), with reported overconsolidation ratios of 1530.
granular column also acts as an effective drainage system to The DBC material is signicantly stiffer and stronger than
prevent excessive buildup of porewater pressure from occur- other well-characterized tills (e.g.  68 times stiffer than
ring (Sivakumar et al., 2013). typical London Clay and  5 times stiffer than typical Cowden
The success of the GA technique for real applications till from the east coast of the UK), at least for the lower strain
requires a method to reasonably predict the loaddisplacement range (Long and Menkiti, 2007; OKelly, 2014). Further
behavior for pullout loading. Various methods of analyses that details on the geotechnical properties and behavior of the
consider different failure modes, including the vertical slip DBC deposit have been reported by Farrell et al. (1995) and
surface model (friction cylinder method) and block type Long and Menkiti (2007). The results of interface shear tests
failures (e.g. inverted cone, circular arc, or in the case of deep on a novel geogrid in DBC backll material have also been
anchors, truncated cone), exist for the determination of the reported by OKelly and Naughton (2008).
ultimate pullout capacity of strip/plate anchors embedded in The BrDBC material is characterized as stiff to very stiff,
uniform deposits of sand/clay (Meyerhof and Adams, 1968; slightly sandy slightly gravelly silt/clay of low plasticity, with
Ilamparuthi et al., 2002; Merield et al., 2001; Merield and typical liquid limit and plastic limit values of 29% and 16%,
Sloan, 2006; Khatri and Kumar, 2009, Rangari et al., 2013). respectively (Long and Menkiti, 2007), and a high bulk unit
Recently, Miyata and Bathurst (2012a, b) investigated the weight of 22 kN/m3 (Kovacevic et al., 2008). Borehole logs for
tensile reinforcement load/pullout capacity of steel strips used the test site indicated that the near saturated BrDBC stratum at
in reinforced soil walls in Japan. However, the failure modes this location was  1.8 m in depth, with a relatively high stone
for GAs are more complex compared with these scenarios; i.e. content (i.e. particle size 4 20 mm) of typically 515% over
strip/plate anchors embedded in uniform deposits of sand/clay. this depth. A very clayey/silty gravel layer was encountered in
This arises on account of the distinctly different response of some of the boreholes at a depth of  0.8 m below ground
the densied gravel material (used to construct the granular surface level (bgl). The standing groundwater table at the site
column) compared with that of the surrounding native mate- was located at between 1.8 and 2.0 m bgl.
rial. For the GA, the applied pullout loading at the ground Fig. 2 shows strength against depth data determined for the
surface is transferred directly to the tendon base-plate assembly test area using a 20 t cone penetration test (CPT) rig and
B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158 1147

Table 1
Anchor installation details.

Anchor Temporary Borehole Anchor Anchor Ultimate eld

number casing diameter, length, L aspect ratio, pullout capacity
required Do (m) (m) L/Do (kN)

GA1 Yes 0.219 1.20 5.5 51.0

GA2 Yes 0.219 0.96 4.4 43.0
GA3 No 0.200 0.50 2.5 19.1
GA4 Yes 0.219 1.00 4.6 47.0
GA5 Yes 0.168 1.47 8.7 42.5
GA6 Yes 0.168 0.80 4.8 33.0
GA7 No 0.150 0.45 3.0 12.8
GA8 Yes 0.168 1.62 9.6 42.0

forming holes greater than 0.5 m in depth for the other GA

Fig. 2. Undrained strength against depth determined from CPT cone-tip installations, the adhesion/friction generated between the fall-
resistance and triaxial compression tests. Note: data labels identify borehole ing cutter tool and sidewalls of the holes was excessive,
numbercored (C)/reconstituted (R) triaxial specimendiameter (mm)applied
cell pressure (kPa).
necessitating the installation of temporary steel casings for
these holes. This had the effect of producing slightly larger
bores with smooth sidewalls. With the casing removed, the
unconsolidatedundrained triaxial compression tests. The bore diameter was the same as the casings outer diameter; i.e.
latter included testing of cored and reconstituted specimens. 168 and 219 mm for hole diameters of nominally 150 and
'Cored' specimens were obtained from just below the base of 200 mm. Into each of these boreholes was placed an M12
the boreholes at nal depth using 38 mm diameter sampling threaded rod (i.e. tendon) with a steel base-plate attachment,
tubes. The reconstituted specimens, 100 mm in diameter 148 and 196 mm in diameters for bores of nominally 150 and
and 200 mm long, were prepared by standard Proctor- 200 mm, respectively. The base plate was secured at the lower
compaction of soil recovered at its in-situ water content using end of the tendon using M12 nuts, one threaded from above
the clay-cutter tool during borehole formation. The CPT the base plate and two threaded from below. The granular
undisturbed undrained shear strength was determined as columns were constructed by backlling uniformly graded

su qc  vo =N kt , where: qc is the cone-tip resistance; vo sub-angular limestone gravel into the boreholes, with compac-
is the overburden pressure and N kt is a cone factor. OKelly, tion to achieve maximum density using the method described
(2014) reported on CPT testing of the DBC deposit at three by Sivakumar et al. (2013). The grading of the gravel (10 mm
different sites in the greater Dublin area. From calibrations nominal particle size) satised the minimum recommended
against measured undrained strengths in triaxial compression, ratio between the nominal particle size and column diameter of
an N kt value of 15 was deemed appropriate for the BrDBC 1:15.
layer and was adopted in the present study. The spiky nature of
the CPT trace is explained by the materials high stone content 2.3. Pullout tests
and occasional gravelly layers/lenses, the presence of which
were conrmed from the recovered cores. From Fig. 2, a Pullout forces were applied to the top ends of the anchor
general trend of increasing strength with depth is evident, with tendons using a hydraulic jack supported above the strong
the remolded undrained shear strength (sur ) at any depth h cross-beam of a reaction frame. For each GA installation, the
given by load against displacement response of the ground-anchor
sur sur0 mh 1 system was measured using a load cell and a displacement
transducer; the latter was mounted on an independent reference
where sur0 is the remolded undrained strength value corre- beam. The vertical displacement of the ground surface was
sponding to ground surface level and m is the rate of strength measured by a second displacement transducer located at a
increase with depth [kPa/m]. For the test area, it was distance of 300 mm from the anchor centerline; i.e. between
determined from Fig. 2 that sur0 64 kPa and m 12.5 kPa/m. 190 and 225 mm (0.87Do 1.5Do ) radially from the sidewalls
of the gravel columns for the different GA installations. The
2.2. Anchor installation displacement response of the ground surface in this region
would be an indicator of the anchors likely failure mechanism,
The 8 anchors (GA1 to GA8, Table 1) were installed in a in that signicant heave would be expected for block type
line of boreholes formed using a light cable-percussion drilling failures or failure in shaft capacity whereas negligible heave would
rig. Boreholes of 150 mm (GA7) and 200 mm (GA3) dia- be expected for GAs failing in end bulging. A single measure-
meters were formed using clay cutter tools. It was found that in ment within this zone was deemed sufcient for this purpose.
1148 B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158

The experimental loaddisplacement and ground heave response 25 50

data are modelled in the second part of this study to better
understand the GAs performance under pullout loading and
associated failure modes. Similar experimental studies performed 20 40
in the future could consider measuring the ground heave response

Ground Heave (mm)


Pullout Force (kN)

at two or more radial distances (each a function of the GAs 15 30
diameter) to provide more experimental data for validation of the
modelling. During application of the pullout load, observations
were made of the relative vertical movements between the tops of 10 20
the gravel columns and the surrounding ground surface. The rate of
loading was such that the anchors ultimate pullout capacity was 5
mobilized within a period of 15 min.

3. Experimental results 0 0
0 50 100 150
Anchor Displacement (mm)
The measured pullout forces and heave of the ground
surface at 0.3 m from the anchor centerline are plotted against
50 5
axial displacement of the anchor tendon (base plate) in Fig. 3.
Visual observations for anchors GA3 and GA7 having GA1(P)
L r 3Do (Fig. 3(a)) indicated that substantial heave of the 40 GA4(P)
surrounding ground occurred on approaching the pullout GA2(P)

Ground Heave (mm)

capacity, with the top surfaces of the gravel columns protrud- Pullout Force (kN) GA8(P) GA6(P)
30 3
ing above the raised ground surface at ultimate pullout
capacity. As expected, a larger column length and/or diameter
produced greater pullout capacity. For longer columns, the 20 2
ultimate pullout capacity was generally mobilized for anchor
displacements of  Do =2; e.g.  85 and  110 mm for GA5 GA2(H)
10 1
(Do 0.168 m) and GA2 (Do 0.219 m), respectively. Even
though displacements of up to 145 mm were required to GA5(H)
mobilize the ultimate pullout capacity of the longest anchors 0 0
(Fig. 3(b)), negligible ground heave (i.e. o 2 mm) was mea- 0 50 100 150
Anchor Displacement (mm)
sured at 0.3 m from the anchor centerline. This suggested that
these anchors had failed in localized bulging near the base of Fig. 3. Experimental values of pullout force and ground heave plotted against
the gravel columns. This was supported by the observation that axial displacement for granular anchors. Note: (P) and (H), pullout force and
heave plots, respectively. (a) L/Do 3. (b) 4.4 L/Do 9.6.
at ultimate pullout capacity, the tops of the gravel columns had
not moved, remaining level with the surrounding ground
in the cavity that forms directly beneath the base plate during
4. Experimental analyses pullout on account of the open pore structure of the gravel
For conventional concrete/steel tension piles, relative dis- Analogous to the analysis of tension piles, for short GAs
placements between the anchor and surrounding ground of failing in shaft capacity, the ultimate pullout load (Pshaf t ) is
 0.5% Do are typically required to mobilize the full shaft given by the summation of the shear resistance mobilized over
capacity. The much larger relative displacements of typically the shaft area and the self-weight of the gravel column (Fig. 4(a)):
 50% Do required to mobilize the ultimate pullout capacities
of the GAs suggested that there were signicant differences Pshaf t Do Lsur L g 2
between the respective load resistance mechanisms. In parti-
cular, one aspect to consider was the signicant increase in where is an adhesion factor; L and Do are the installed
lateral connement pressure induced on the granular column (initial) column length and diameter respectively; sur is the
during pullout loading on account of the dilation of the dense mean remolded undrained strength over the column length and
gravel. g is the unit weight of gravel forming the granular column.
An undrained analysis was justied for the surrounding From Eq. (1) and Fig. 2, sur 6774 kPa for the 8 GAs
soil considering: (a) intact BrDBC material has a (horizontal) reported in the present study. As described earlier in the paper,
permeability coefcient value of the order of 10  9 m/s (Long the borehole formation process generally required a temporary
and Menkiti, 2007); and (b) the GAs ultimate capacities were steel casing which had the effect of produced a smooth bore
mobilized within 15 min of starting the pullout tests. Note that sidewall. Under vertical loading, conned compression of the
for the experimental setup described, a vacuum cannot develop gravel column and dilation of the dense gravel accompanying
B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158 1149

Fig. 4. Mobilization of resistance in GAs under pullout loading (a) Failure in shaft capacity (L o ~ 6Do, Eq. (2)). (b) Small force resisted in shaft resistance over
lower section of long column. (c) Shaft resistance mobilizing upwards along column to resist increasing load. (d) Failure in localized end bulging of column (L o ~
6Do, Eq. (3)). (e) Encasement of lower section of gravel column to impose failure in shaft capacity.

the large relative displacements between the GA and surround- compression (Eq. (3)). Localized bulging for stone columns
ing soil produced signicant increases in the normal stresses under compression loading and long GAs under pullout loading
acting at the soilcolumn interface. Under these conditions, occurs because of lack of sufcient lateral connement at the
some embedment of the gravel particles into the bore sidewall top and bottom ends, respectively, of the granular columns.
was inevitable. Hence, at ultimate pullout capacity, the rupture D2 vbase
surface occurs within the soil next to the column shaft. Pbase 3
Signicant remolding occurs within this zone on account of
where D is the diameter of the column bulge; vbase is the
the borehole formation process and the large relative displace-
bearing pressure at the column base which is  estimated by
ments occurring between the column shaft and surrounding
vbase 1 sin 0g =1  sin 0g  vc N nc surbase , in which 0g
soil during pullout loading. Under these circumstances, an
is the gravels effective friction angle; N nc is a bearing capacity
value of unity is appropriate, as demonstrated by Sivakumar
factor considering local shear failure; vc is the overburden
et al. (2013) from back analysis of the eld performance of
pressure provided by the surrounding ground and surbase is the
GAs installed in aged made ground deposits.
remolded undrained strength in the bulging zone.
For longer GAs, an increasing uplift force applied by the
The local bearing capacity factor is given by Gibson and
anchor tendon to the base plate is rst resisted in shaft
Anderson (1961):
resistance over the lower section of the gravel column
(Fig. 4(b)). The relative movements between the column and N nc 1 log 4
surrounding soil mean that the shaft resistance initiates from surbase
the column base and develops upwards along the column where Gu is the undrained shear modulus.
length. As the applied force increases further, shaft resistance The overburden pressure is given by vc s L0 , where s is
is mobilized over an increasing distance from the column base the bulk unit weight of the surrounding soil and L0 is the
(Fig. 4(c)), up to a point when structural failure of the gravel overburden depth to the mid-height of the bulge zone.
column occurs by localized end bulging because of a lack of Sivakumar et al. (2013) suggested that a localized enlargement
sufcient lateral connement in the immediate vicinity of the of approximately 10% in the column diameter occurred on
highly stressed column base (Fig. 4(d)). With the buildup in nearing failure in end bulging; i.e. in Eq. (3), D E 1.1Do .
end bulging resistance of the column (accompanied by large Assuming no signicant movement of the gravel material
localized strains), the mobilized shaft resistance reduces back. occurs above the bulging zone and conservation of volume for
In other words, the dominant failure mode is governed by the the dense gravel, it can be determined that the predicted length
columns L=Do ratio. of the bulge zone at pullout failure (typically occurring for
For GAs failing in end bulging, Sivakumar et al. (2013) axial displacements of  Do =2) is  2.5Do . Hence the mid-
suggested that the ultimate capacity Pbase can be determined by height of the bulge zone at ultimate pullout capacity occurs for
adapting the method presented by Hughes et al. (1975) for an overburden depth of L0 EL  Do 2:5Do =2 L  1:75Do
calculating the ultimate capacity of stone columns under (see Fig. 4(d)).
1150 B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158

40 Table 2
Shaft capacity (Eq. 2) Material parameter values.

Material Value
End bulging capacity (Eq. 3) GA5 GA8
Surrounding soil
Bulk unit weight, s (kN/m3) 22
GA6 Remolded undrained strength at ground surface level, sur0 (kPa) 64

20 GA1
GA4 Rate of increase in undrained strength with depth, m (kPa/m) 12.5
GA2 Undrained Youngs modulus at ground surface, Euo50 (MPa) 7.0
Rate of increase of Youngs modulus with depth, Euo50 (MPa/m) 1.4
10 GA7 Undrained Poissons ratio, u 0.5
Coefcient of earth pressure at rest, K0 1.5
Gravel column
0 Bulk unit weight, g (kN/m3) 20
0 2 4 6 8 10 12 Apparent cohesion, c0 (kPa) 0.2
L/D Effective friction angle, 0 g (degree) 42
Dilatency angle, 0 (degree) 10
Fig. 5. Non-dimensional ultimate pullout capacity against column L=D ratio Drained Youngs modulus at ground surface level (MPa) 4.5
for granular anchors. Rate of increase in Youngs modulus with depth (MPa/m) 30
Drained Poissons ratio, 0 0.3

The ultimate pullout load in shaft capacity increases

proportionally with, and is strongly sensitive to, the columns depth of 0.80.9 m bgl at the location of anchor GA6. Its
L=D ratio. Above a critical aspect ratio L=Do cr , failure in presence can also be inferred from the signicantly higher CPT
end bulging is the dominant mechanism, with the GAs cone-tip resistance values mobilized over this depth range (see
capacity dependent on G=surbase , 0g and its L=D ratio (see Fig. 2). This would explain why the measured ultimate pullout
Eq. (3)). As shown later in the paper, for a given column capacity of GA6 was greater than its shaft capacity predicted
diameter, the ultimate pullout capacity for failure in end using the representative soil property values, reported in
bulging increases only marginally with increasing L=D ratio. Table 2. All four anchors of 200 mm nominal diameter had
Fig. 5 shows the experimental ultimate pullout capacity L=D r 5.5 (see Table 1), indicating that they had failed in shaft
values for the 8 GAs, expressed in the non-dimensional form capacity. By contrast, anchors GA5 and GA8 (L=D of 8.7 and
of Pn ( 4Pmeasured =D2o sur ), plotted against the columns L=D 9.6, respectively) failed in end bulging. The hypothesis was
ratios. Also included in this gure are envelopes of ultimate substantiated by the insignicant heave (r 0.15 mm, Fig. 3
resistance in shaft capacity and in end bulging predicted using (b)) of the ground surface measured at 0.3 m from the
Eqs. (2) and (3), respectively, but expressed in the form of centerline of these two anchors at ultimate pullout capacity.
Pn shaf t (=4L=Do L g =sur ) and Pn base (= vbase =surbase ). An
value of unity (Sivakumar et al., 2013) was used in computing 5. Numerical analyses
the shaft capacity values. The supposed transition between the
different failure modes for the specic ground conditions The numerical analyses were performed using a commer-
encountered at the test site occurred for L=Do cr E6.2. The cially available nite-element program (PLAXIS 2D 2010
pertinent soil parameter values used in these calculations are (Brinkgreve et al., 2010)), employing 15 node triangular
listed in Table 2. elements and invoking axisymmetry. The BrDBC material
Since the GAs had been quickly loaded to failure, with the was modeled using a total stress approach (su ,u 0), con-
surrounding soil remaining in an undrained condition, the sistent with the experimental conditions. Furthermore, all of
BrDBCs shear modulus value for computing the local the soil parameter values measured were for the undrained
bearing-capacity factor N nc in Eq. (4) could be estimated using condition. The gravel columns were modeled using an
elastic theory, with an undrained Poissons ratio (u ) value of effective stress approach. A MohrCoulomb model was used
0.5. However good-quality undisturbed sampling of the for the BrDBC and gravel materials, with consideration of the
BrDBC layer was not possible on account of its high stone increase in undrained strength and stiffness with depth. The
content. Hence, in the present investigation, a single opera- use of the MohrCoulomb model for the BrDBC layer was
tional Gu value of 3.0 MPa was assumed for the BrDBC layer, justied since this material is highly overconsolidated, with
and based on the mean surbase value of  77 kPa determined reported overconsolidation ratio values ranging 1530. A
for the 8 GAs tested, an N nc value of 4.7 is obtained using Eq. typical apparent pre-consolidation (yield) stress value of
(4).  1.0 MPa was estimated from the corrected CPT cone-tip
Deviations between the experimental and predicted pullout resistance (qt ) data using the method after Kulhawy and
capacity values presented in Fig. 5 most likely occurred on Mayne (1990). This apparent pre-consolidation stress for the
account of the inherent variability/strength heterogeneity of the test site is in general agreement with the value of 750 kPa for
BrDBC layer at the test site. For instance, a very clayey/silty BrDBC determined from in-situ dilatometer tests reported by
gravel layer had been conrmed from the borehole risings for a Lawler et al. (2011).
B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158 1151

The Youngs modulus values adopted in the numerical Dublin have assumed K0 values for the BrDBC layer ranging
analyses required special attention. When using a constant 1.01.5 in design (Long and Menkiti, 2007). Based on this
stiffness modulus to represent soil behavior (as in the Mohr evidence, a constant K 0 value of 1.5 with depth was adopted in
Coulomb model), one should choose a value that is consistent the present study. For numerical reasons, an undrained
with the stress level and stress path development. The pertinent Poissons ratio value of 0.495 was employed along with an
input parameters are values of undrained (secant) Youngs apparent cohesion c0 value of 0.2 kPa for the gravel.
modulus at 50% shear strength corresponding to ground An axisymmetric model with standard xities and dimen-
surface level (E uo50 ) and the rate of increase in this modulus sions of 2.5 m in radius and 2.5 m in depth was used for all of
with depth (E uo50 ). Both values relate to a reference conn- the simulations. This placed the outer vertical boundary at a
ing pressure of 100 kPa in the triaxial cell since their values distance of at least 11Do from the sidewall of the gravel
tend to increase with conning pressure. Since undisturbed column and allowed freedom for any of a number of possibly
samples were not available, a different approach was adopted mechanisms to develop in the BrDBC material, without
in the determination of these stiffness values. Twelve triaxial signicant inuence from the outer boundary. As for the in-
specimens, each 100 mm in diameter and 200 mm long, were situ condition, the phreatic level was set at 1.8 m bgl.
prepared by standard Proctor-compaction of BrDBC material The calculation scheme was performed in three stages: (a)
that had been recovered at its natural water content from the initial stresses were generated in the 2.5 m thick BrDBC
different depths using the clay cutter tool during borehole layer using the K 0 procedure; (b) the GAs gravel column was
formation. These specimens were tested in unconsolidated wished-in-place; (c) the operation of the anchor during
undrained triaxial compression, with the stiffness values at pullout loading (i.e. uniform upward movement of its rigid
50% shear strength determined from the measured stressstrain base plate) was simulated by means of an upward prescribed-
curves. The values of E uo50 7.0 MPa and Euo50 1.4 MPa/ displacement condition acting over the base of the gravel
m depth were deduced from regression analysis of the stiffness column. The horizontal dimension (width) of the prescribed
values at 50% shear strength plotted against depth for the 12 displacement was set equal to that of the base plates used in the
triaxial specimens. It is acknowledged that this approach eld tests, simulating the initial gap of  10 mm present
cannot reproduce the inherent structure of the ground and between the outer rim of the base plate and the bore sidewall.
may result in (signicantly) lower values of soil stiffness, A tension cutoff value of 0 kPa was specied throughout the
especially at small strains. With mean values of L E1.0 and BrDBC layer; i.e. vacuum cannot develop in the cavity that
surbase E77 kPa for the 8 GAs tested, these stiffness values forms directly beneath the base plate during pullout. A number
indicate Gu E2.8 MPa (from Gu Eu =3), which is consistent of simulations performed for different mesh densities indicated
with the value of 3.0 MPa adopted for the BrDBC layer in the that coarse meshing (with approximately 1100 elements) was
experimental analyses. For the drained Poissons ratio of 0.2 adequate, with pullout failure typically achieved within
reported for BrDBC (Kovacevic et al., 2008), the E uo50 and 5000 steps.
Euo50 values used in the numerical analyses correspond to Simulations were also performed for a modied base-plate
drained modulus values of 5.6 MPa and 1.1 MPa/m depth, arrangement that allowed suctions of up to one atmosphere to
respectively. develop in the cavity formed beneath the base plate during
Considering the very low connement pressure, a relatively pullout. This condition could occur for (near) saturated, low
low drained Youngs modulus of 4.5 MPa was adopted at permeability soils under relatively quick applied loading. Such
ground surface level for the dense gravel column. Its value was an anchor arrangement could involve an inverted cup (bucket)
considered to increase signicantly and proportionately with attachment at the bottom end of the tendon, which would be
depth. The 0g value of 421 adopted is consistent with reported driven (embedded) into the base of the borehole (Fig. 6(a)).
peak values for dense sub-angular gravel. This scenario was modeled by specifying a tension cutoff
The MohrCoulomb model applied in PLAXIS 2D (2010) value of 100 kPa for the BrDBC material. Such an arrange-
does not allow for dilatency cutoff; i.e. end of dilatency occurs ment could also mitigate against the tendency for plastic ow
when the soil reaches the critical state. The effect of dilatency of soil from the bulge zone into the cavity forming at the
angle 0 was investigated by running simulations with input 0 column base by the upward movement of the anchor (Fig. 6
values of 101 and then 51; i.e. moving towards the critical state (b)).
0 01 value. The interactions between the gravel and BrDBC
materials in contact with the top and bottom surfaces, 6. Numerical results
respectively, of the base plate were modeled using an interface
friction coefcient value of 0.67. Fig. 7 shows predicted GA pullout resistances along with
Long and Menkiti (2006, 2007) and Lawler et al. (2011) ground heave responses at 0.3 m from the anchor centerlines.
reported an average coefcient of earth pressure at-rest (K0) Good overall agreement was achieved between the measured
value of 1.5 for the BrDBC layer, determined from high and predicted values of ultimate pullout capacity and the
quality in-situ dilatometer tests. In previous nite element corresponding anchor (base plate) displacements. Deviations
analyses, values of K0 1.5 (Menkiti et al., 2004; Kovacevic between the measured and predicted pullout forces arose due
et al., 2008) and 3.0 (Lawler et al., 2011) have been adopted to the inherent variability/strength heterogeneity of the BrDBC
for the BrDBC layer. In the absence of data, engineers in layer over the test area, with the simulations performed using
1152 B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158

Fig. 6. Outline of modied base-plate arrangement for improved GA performance. (a) Proposed installation. (b) At pullout failure in shaft capacity.

representative soil parameter values. Another factor was the and GA8 (L=D of 4.6 and 9.6, respectively). For GA8 (Fig. 9
material model adopted, with the MohrCoulomb (linear- (b)), no increase in normal stress was predicted over the upper
elastic perfectly plastic) representation used for the gravel half of the column length. This can be explained by referring to
column and surrounding soil predicting a stiffer response for Fig. 4(bd). Under upward displacement of the base plate
the groundanchor system and substantially overestimating the caused by increasing pullout load, conned compression of the
ground heave, particularly for experimental GAs having gravel column and dilation of the dense gravel produces some
L=D r 5.5 [i.e. o L=Do cr ]. For GA5 and GA8 (L=D Z 8.7), embedment of the gravel particles into the bore sidewall and a
the measured and predicted ground heave responses were in buildup in normal stress that propagates upwards from the
reasonable agreement, signicantly smaller in magnitude and column base. The pullout load is resisted in shaft capacity
approximately increased in proportion with the anchor dis- mobilized over this lower section of the column until such
placements. Again, the distinctly different ground heave point that the normal stresses become too great, resulting in
responses for experimental anchors having L r 5.5Do and localized end-bulging failure. In this scenario, no increase in
Z 8.7Do indicated different failure mechanisms were at play. normal stress or relative movement (and hence shaft resistance
Fig. 8 shows the extent of the plastic zones predicted in the development) occurs over the upper section of the column
soil surrounding the GAs at ultimate capacity. From these, the length. By contrast, for GA4 (Fig. 9(b)), the normal stresses
different failure mechanisms occurring predominantly in shaft increased and relative movements occurred at the interface for
capacity (Fig. 8(ac)) or in end bulging (Fig. 8(d)) can be the full column length, indicative of full mobilization of the
deduced and are dependent on the column L=D ratio. The shaft capacity.
enlarged plastic zone formed near the base of anchor GA8 Fig. 10 shows the radial expansion of the bore sidewall
(L=D 9.6, Fig. 8(d)) is indicative of failure in end bulging, predicted for different depths (characterized by values of z=Do,
consistent with measured and predicted ground heave move- where z is the distance measured from the column base) along
ments and also with the experimental analyses presented the lower section of the gravel column. Fig. 10(a and b) shows
earlier. For all GAs tested having L r 5.5Do , plastic zones negligible radial expansion of the gravel columns was pre-
developed over the full column length in the soil next to the dicted for GAs having L=D r 3.0. Radial strains r (computed
soilcolumn interface (conrmed by contours of displacement as the radial expansion expressed as a percentage of the GAs
plots), indicative of failure in shaft capacity. The extent of the initial column radius) of less than 2.1% were predicted for the
tension zones at the ground surface extended to  1.5 m anchor displacements ( 45 mm, Fig. 3(a)) corresponding to
( 7Do ) from the anchor centerline. the eld ultimate pullout capacity. However, for GA5 and GA8
Fig. 9 shows contours of normal (radial) stress predicted (L=D Z 8.7, Fig. 10(g and h)), signicant bulging of the
over the column length at ultimate pullout capacity for GA4 columns was predicted over a length of  23Do from the
B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158 1153

30 80 60 40
Measured load (ML)
Predicted load (PL)
With suction cup (SC)
Measured heave (MH) 60 30

Ground Heave (mm)

Ground Heave (mm)
Pullout Force (kN)

Pullout Force (kN)

Predicted heave (PH) PH
20 40

40 20

10 20 Measured load (ML)

PH Predicted load (PL)
20 With suction cup (SC) 10
Measured heave (MH)
MH Predicted heave (PH)
0 0 0 0
0 20 40 60 80 100 0 50 100 150 200
Anchor Displacement (mm) Anchor Displacement (mm)

Measured load (ML) Measured load (ML)

Predicted load (PL) Predicted load (PL)
80 With suction cup (SC) 20 60 With suction cup (SC) 5
Predicted load, ' = 5o (PL(5o)) Predicated load, ' = 5o (PL(5o))
Measured heave (MH) Measured heave (MH) SC
Predicted heave (PH) SC Predicted heave (PH) 4

Ground Heave (mm)

60 15 Ground Heave (mm)
Pullout Force (kN)

Pullout Force (kN)

PL(5o) 40 PL
40 10
PH 20 ML
20 5

0 0 0 0
0 50 100 150 200 0 50 100 150 200
Anchor Displacement (mm) Anchor Displacement (mm)

Fig. 7. Predictions of pullout resistance and ground heave at 0.3 m from the anchor centreline plotted against anchor displacement, with the plots ordered by
increasing column L=D ratio. Unless otherwise stated, simulations are for a constant 0 101. (a) GA7 (L/D=3.0). (b) GA2 (L/D = 4.4). (c) GA1 (L/D = 5.5). (d)
GA8 (L/D = 9.6).

column base, with r values of  35% predicted for the much the value of r E 10% postulated by Sivakumar et al. (2013)
larger anchor displacements of at least 100 mm require to for failure of the gravel column in end bulging. This is most
mobilize eld ultimate pullout capacity (Fig. 3(b)). For likely explained by the overestimation of the dilatancy for the
intermediate L=D values, some radial expansion of the gravel gravel in the numerical predictions (which were based on a
column was also predicted to occur within 23Do from the constant 0 101), whereas 0 01 at critical state. In other
column base; e.g. r 814% for the anchor displacements words, in the numerical analyses, the ultimate pullout capacity
corresponding to the eld ultimate pullout capacity of GAs 1, 2 and corresponding ground heave movements for these anchors
and 4. However, this r range is not enough to develop were overestimated. This is conrmed by comparing Figs. 10(g
sufcient bulging resistance for failure to occur in end bulging. and h) and 11(a and b), with predicted r values reducing by
 12% when the input dilatency angle (which remains xed
6.1. Length of the bulge zone throughout the numerical simulation) was reduced from 101 to 51.
Fig. 12 shows non-dimensional ultimate pullout capacity
For GAs failing predominantly in end bulging at the test site (Pn ) predictions for the 8 GAs plotted against column L=D
(i.e. L=D Z 6.2), the predicted bulge length of  23Do is ratio. The predicted Pn values for GAs failing in shaft capacity
consistent with the value of  2.5Do determined earlier using (i.e. L=D o 6.2) were in good agreement with the trend line
assumptions reported by Sivakumar et al. (2013) regarding end given by Eq. (2), but expressed in non-dimensional form.
bulge formation. Some bulging of the gravel columns was also However, for anchors GA5 and GA8 failing in end bulging
predicted at distances of up to  8Do from the column base, (L=D Z 8.7), the predicted bulging capacities overestimated
although its amount reduced signicantly with decreasing the bulge trend line given by Eq. (3), expressed in non-
depth over this zone. dimensional form. This can be explained by the constant 0
The r values of  35% and predicted for the anchor value of 101 used in these numerical simulations. Since the
displacements corresponding to the eld ultimate pullout dilatency angle is not explicitly considered in Eq. (3), the
capacities of GA5 and GA8 were signicantly greater than agreement between the experimental data and the bulge trend
1154 B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158

Fig. 8. Extent of plastic zone predicted at ultimate pullout capacity for a constant 0 101. Note: MohrCoulomb points and tension-cutoff points are indicated by
red shading and hollow boxes, respectively. Black dotted lines dene extents of tension cutoff zones. (a) GA7 (L/D = 3.0). (b) GA4 (L/D = 4.6). (c) GA1 (L/D =
5.5). (d) GA8 (L/D = 9.6). (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

line was good (Fig. 5). In practice, however, with large benet achieved for GAs failing in end bulging was minor,
localized deformations occurring during column end-bulging, with negligible improvement achieved for L=D Z 10. Further
the 0 value for the gravel reduces towards the critical state investigations and validation using experimental eld trials are
0 01 value. In order to validate this hypothesis, a number of necessary to conrm these ndings.
the simulations were repeated using a lower (constant) 0 value
of 51 (e.g. see Fig. 7(c and d)), which was found to produce 7. Discussion
much better agreement with Eq. (3) trendline (see Fig. 12).
Using experimental and numerical means, this paper has
6.2. Modied anchor base-plate for improved pullout capacity conrmed that failure of GAs predominantly occurs in shaft
capacity or in end bulging, depending on the columns L=D
Fig. 12 demonstrates the effect of developing suction of one ratio. Setting Pshaf t Pbase (Eqs. (2) and (3), respectively) and
atmosphere in the cavity that forms directly beneath the base disregarding the small contribution of the columns self-weight
plate during pullout loading (see 'With suction cup' data in component (i.e. second term in Eq. (2)), the transition between
gure). The predicted improvement in ultimate pullout capa- failure in shaft capacity and in end bulging occurs for
city was found to decay exponentially with the column L=D
Lcr D vbase
ratio (Fig. 13). From the numerical analyses, the proposed 5
modication of the base-plate arrangement produced signi- Do 4 sur
cant increases in the undrained ultimate pullout capacity for with vbase and hence Lcr =Do dependent on 0g , surbase and Gu.
short GAs; e.g. between  30% (L 2.5Do ) and 6% Note that the value of Lcr =Do increases signicantly with 0g ,
(L 6.2Do ) for GAs failing in shaft capacity. However the but only marginally with the Gu=surbase ratio.
B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158 1155

However, once the column is fully encased for depths greater

than  6Do , the hoop resistance provided will prevent loca-
lized bulging failure from occurring. Hence, under increasing
applied pullout loading, the shaft resistance can continue to
develop upwards to the top of the gravel column (Fig. 4(e)),
with failure eventually occurring exclusively in shaft capacity.
The numerical analysis has shown that the undrained ultimate
pullout capacity can be signicantly increased for short GAs
installed in (near) saturated, low permeability soils by using an
inverted cup (bucket) in place of the conventional at anchor
base plate.
Finally, all of the eld tests and numerical simulations
presented in this paper relate to the pullout capacity mobilized
for the undrained condition. Hence the potential for some
softening/swelling of the soil in the vicinity of the column
base/bulge zone (e.g. as a result of the groundwater regime or
surface water entering down the column shaft) could cause
some reduction in the ultimate pullout capacity, particularly for
over-consolidated clays.

8. Conclusions

Using experimental and numerical means, this paper has

conrmed that the undrained pullout capacity of granular
anchors (GAs) is mobilized in shaft capacity or in end bulging,
depending on the columns L=D ratio. During pullout loading,
conned compression of the column and dilation of the dense
gravel under the large relative displacements occurring at the
soilcolumn interface produce signicant increases in the
normal stresses and hence some embedment of the gravel
particles into the sidewall of the soil bore. For GAs failing in
Fig. 9. Predicted normal stress contours (in red color) at ultimate pullout shaft capacity, the rupture surface occurs within the remolded
capacity for a constant 0 101. (a) GA4 (L/D = 4.6). (b) GA8 (L/D = 9.6). soil next to the column shaft, with the ultimate pullout capacity
(For interpretation of the references to color in this gure legend, the reader is
increasing strongly and proportionally with the column L=D
referred to the web version of this article.)
ratio. At the ground surface, the extent of the tension zone in
the surrounding soil extends a distance of  7Do from the
anchor centreline. Above a critical column aspect ratio
For the particular soil conditions at the test site, the (L=Do ) cr value, at ultimate pullout capacity the column fails
transition between the two failure modes occurred for structurally by bulging over its lower end (concentrated at
L=Do cr E 6.2. This value is consistent with experimental  23Do from the column base), with its capacity dependent
observations from other full-scale pullout tests reported for on Gu =surbase , 0g and the column L=D ratio. The eld ultimate
GAs by OKelly et al. (2013) and Sivakumar et al. (2013). pullout capacity for end bulging failure was substantially
Numerical predictions of the bulge formation, concentrated mobilized for anchor displacements of  Do =2 and increases
within a region extending to 23Do from the column base, are only marginally in value with increasing L=D ratio. For the
also consistent with assumptions reported by Sivakumar et al. particular ground (intact lodgement till) at the tests site and
(2013). granular backll material used to form the columns, the
Several researchers (e.g. Phani Kumar and Ramachandra transition between the two failure modes occurred for (L=Do )
Rao (2000) and Sharma et al. (2004)) have reported that end cr E6.2. The value of (L=Do ) cr increases signicantly with 0g
bulging failure of long GAs can be contained by encasing the and marginally with Gu =usurbase . Numerical analyses also
lower section of the gravel column with geotextile (geofabric showed that the undrained ultimate pullout capacity can be
tube/sock), thereby providing better performance; i.e. ultimate increased (signicantly for short GAs) by using an inverted cup/
pullout capacity increases and tendon displacements under bucket in place of the at base-plate arrangement used in
pullout loading decrease. The encasement of the lower section previous GA setups. The benet of the proposed modication
of the gravel column would tend to push the zone of bulging decayed exponentially with increasing L=D ratio, with no
higher up the column, where the conning stresses are lower. signicant gain achieved for LZ10Do .
1156 B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158

Fig. 10. Predicted radial expansion of gravel column for different z=Do; where z is the distance from the column base. Unless otherwise stated, simulations are for a
constant 0 101. (a) GA3 (L/D = 2.5). (b) GA7 (L/D = 3.0). (c) GA2 (L/D = 4.4). (d) GA4 (L/D = 4.6). (e) GA6 (L/D = 4.8). (f) GA1 (L/D = 5.5). (g) GA5 (L/D
= 8.7). (h) GA8 (L/D = 9.6).
B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158 1157


Increase in ultimate pullout capacity (%)

Lateral Expansion (mm)

30 30

20 2D

10 3D
0 50 100 150 200
Anchor Displacement (mm) 0
0 2 4 6 8 10 12
Fig. 13. Predicted increase in ultimate pullout capacity for suction of one
atmosphere developed beneath the anchor base plate (assuming constant
0 101 for gravel column).
Lateral Expansion (mm)


Trinity College Dublin) for assistance in performing the eld

20 2D and laboratory tests presented in this paper. The numerical
analyses part of the paper was performed by the lead author at
PLAXIS bv, The Netherlands, during a period of sabbatical
3D leave from Trinity College Dublin. This element of the
z=D research was performed under the Marie Curie IAPP project
8D NOTES (Grant no. PIAP-GA-2008-230663).
0 50 100 150 200
Anchor Displacement (mm)
Fig. 11. Predicted radial expansion of gravel column for a constant 0 51. (a) Brinkgreve, R.B.J., Swolfs, W.M., Engin, E., 2010. PLAXIS 2D 2010.
GA5 (L/D = 8.7). (b) GA8 (L/D = 9.6). PLAXIS bv, The Netherlands.
Farrell, E.R., Coxon, P., Doff, D.H., Priedhomme, L., 1995. The genesis of
the brown boulder clay of Dublin. Q. J. Eng. Geol. 28 (2), 143152.
Gibson, R.E., Anderson, W.F., 1961. In-situ measurements of soil properties
with the pressuremeter. Civ. Eng. Public Works Rev. 56 (658), 615618.
Hughes, J.M.O., Withers, N.J., Greenwood, D.A., 1975. A eld trial of the
reinforced effect of stone column in soil. Gotechnique 25 (1), 3144.
Ilamparuthi, K., Dickin, E.A., Muthukrisnaiah, K., 2002. Experimental
investigation of the uplift behaviour of circular plate anchors embedded
in sand. Can. Geotech. J. 39 (3), 648664.
Khatri, V.N., Kumar, J., 2009. Vertical uplift resistance of circular plate
anchors in clays under undrained condition. Comput. Geotech. 36 (8),
Kovacevic, N., Milligan, G.W.E., Menkiti, C.O., Long, M., Potts, D.M., 2008.
Finite element analyses of steep man-made cuts in Dublin boulder clay.
Can. Geotech. J. 45 (4), 549559.
Kulhawy, F.H., Mayne, P.H. 1990. Manual on Estimating Soil Properties for
Foundation Design. Electric Power Research Institute (EPRI), Palo Alto
CA. Report EL-6800.
Lawler, M.L., Farrell, E.R., Lochaden, A.L.E., 2011. Comparison of the
measured and nite element-predicted ground deformations of a stiff
lodgement till. Can. Geotech. J. 48 (1), 98116.
Fig. 12. Pn predictions against column L=D ratio.
Liu, K.F., Xie, X.Y., Zhang, J.F., Zhu, X.R., 2006. Compression/tension load
capacity of stone column anchors. Proc. Inst. Civ. Eng. Geotech. Eng.
Acknowledgements 159 (3), 161165.
Long, M., Menkiti, C.O., 2007. Geotechnical properties of Dublin Boulder
Clay. Gotechnique 57 (7), 595611.
The authors thank Martin Carney and Eoin Dunne (Depart- Long, M., Menkiti, C.O. 2006. Characterisation and engineering properties of
ment of Civil, Structural and Environmental Engineering, TCD Dublin Boulder Clay. In: T.S. Tan, K.K. Phoon, D.W. Hight, S. Leroueil
1158 B.C. OKelly et al. / Soils and Foundations 54 (2014) 11451158

(Eds.), Proceedings of the Second International Workshop on Character- Las Vegas, Nevada, USA, Paper number 2-50, 8 pp. http://www.cpt14.
isation and Engineering Properties of Natural Soils Taylor and Francis, com/cpt14-papers,
London, UK, vol. 3, pp. 20032045. Phani Kumar, B.R., Ramachandra Rao, N., 2000. Increasing pull-out capacity
Miyata, Y., Bathurst, R.J., 2012a. Analysis and calibration of default steel strip of granular pile anchors in reactive soils using base geosynthetics. Can.
pullout models used in Japan. Soils Found. 52 (3), 481497. Geotech. J. 37 (4), 870881.
Miyata, Y., Bathurst, R.B., 2012b. Measured and predicted loads in steel strip Phanikumar, B.R., Sharma, R.S., Srirama Rao, A., Madhav, M.R., 2004.
reinforced c soil walls in Japan. Soils Found. 52 (1), 117. Granular pile anchor foundation (GPAF) system for improving the
Menkiti, C.O., Long, M., Kovacevic, N., Edmonds, H.E., Milligan, G.W.E., engineering behaviour of expansive clay beds. Geotech. Test. J. 27 (3),
Potts, D.M., 2004. Trial excavation for cut and cover tunnel construction in 19.
glacial till: a case study from Dublin. In: R.J. Jardine, D.M. Potts, Phanikumar, B.R., Srirama Rao, A., Suresh, K., 2008. Field behaviour of
K.G. Higgins, (Eds.), Skempton Memorial Conference: Advances in Geotech- granular pile-anchors in reactive soils. Proc. Inst. Civ. Eng. Ground
nical Engineering, Thomas Telford, London, UK, vol. 2, pp. 10901104. Improv. 161 (4), 199206.
Merield, R.S., Sloan, S.W., Yu, H.S., 2001. Stability of plate anchors in Rangari, S.M., Choudhury, D., Dewaikar, D.M., 2013. Seismic uplift capacity
undrained clay. Gotechnique 51 (2), 114153. of shallow horizontal strip anchor under oblique load using pseudo-
Merield, R.S., Sloan, S.W., 2006. The ultimate pullout capacity of anchors in dynamic approach. Soils Found. 53 (5), 692707.
frictional soils. Can. Geotech. J. 43 (8), 852868. Sharma, R.S., Phanikumar, B.R., Nagendra, G., 2004. Compressive load
Meyerhof, G.G., Adams, J.I., 1968. The ultimate uplift capacity of foundations. response of granular piles reinforced with geogrids. Can. Geotech. J. 41
Can. Geotech. J. 5 (4), 225244. (1), 187192.
OKelly, B.C., Naughton, P.J., 2008. On the interface shear resistance of a Sivakumar, V., OKelly, B.C., Madhav, M.R., Moorhead, C., Rankin, B.,
novel geogrid with in-plane drainage capability. Geotext. Geomembr. 26 2013. Granular anchors under vertical loading/axial pull. Can. Geotech. J.
(4), 357362. 50 (2), 123132.
OKelly, B.C., Sivakumar, V., Brinkgreve, R.B.J., 2013. Uplift resistance of Srirama Rao, A., Phanikumar, B.R., Dayakar Babu, R., Suresh, K., 2007.
granular anchors in clay under undrained condition. In: Proceedings of Pullout behavior of granular pile-anchors in expansive clay beds in situ.
the First Workshop on SoilStructure Interaction organised by the Marie Geotech. Geoenviron. Eng. 133 (5), 531538.
Curie IAPP project NOTES, Trinity College Dublin, Dublin, Ireland. Xu, H.-y., Chen, L.-z., Deng, J.-l., 2014. Uplift tests of jet mixing anchor pile. Soils Found. 54 (2), 168175.
OKelly, B.C., 2014. CPT testing in Dublin Boulder Clay. In: Proceedings of
the Third International Symposium on Cone Penetration Testing (CPT14),